About this example
This page is an example methodology chapter for a doctoral thesis built around a systematic review and meta-analysis in oncology. In a review-based thesis, the methodology chapter is essentially the review protocol: it fixes the question, the search, the selection rules, the risk-of-bias assessment and the synthesis before any results are seen. The example shows what a complete protocol contains and why each element is there.
It is an illustrative teaching example, not a real student's work, and it reports no results. The clinical question is deliberately framed in general terms and should be refined with an oncology supervisor. Read each part, then the "What to notice" note beneath it. Do not submit any part of it as your own work; your university's academic integrity rules apply.
The example: review question and registration
3.1 Review design
This is a systematic review of intervention effects, with meta-analysis where studies are sufficiently similar (Grant & Booth, 2009). It will be reported according to PRISMA 2020 (Page et al., 2021) and was registered on PROSPERO before screening began; any changes to this protocol will be dated and explained in the thesis.
3.2 Review question
| Element | Definition in this review |
|---|---|
| Population | Adults (18 years or over) with advanced or metastatic solid tumours |
| Intervention | Immune checkpoint inhibitors, alone or combined with other systemic therapy |
| Comparator | Standard systemic therapy without a checkpoint inhibitor, as defined in each trial |
| Outcomes | Primary: overall survival. Secondary: progression-free survival, grade 3 or higher adverse events, health-related quality of life |
| Study designs | Randomised controlled trials; non-randomised studies only for long-term safety |
The example: search and selection
3.3 Information sources and search
MEDLINE, Embase, the Cochrane Central Register of Controlled Trials (CENTRAL) and Web of Science will be searched from database inception, because no single database retrieves all relevant studies (Bramer et al., 2017). ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform will be searched for completed trials with posted results, and conference abstracts from the past five years will be screened, because excluding unpublished and grey literature can bias the estimates of meta-analyses (Hopewell et al., 2007). The strategy combines controlled vocabulary and free-text terms for the population, the intervention class and randomised designs; the full MEDLINE strategy is given in Appendix A. No language limits will be applied.
3.4 Selection and data extraction
Two reviewers will independently screen titles and abstracts and then full texts against the criteria in Table 3.1, resolving disagreements by discussion or with a third reviewer. Reasons for excluding full texts will be recorded for the PRISMA flow diagram. Data will be extracted independently by two reviewers into a piloted form covering study design, participants, interventions, comparators, outcome definitions, follow-up and results, including hazard ratios with confidence intervals for time-to-event outcomes.
Oncology trials are often reported several times, as interim and final analyses, conference abstracts and secondary publications. Reports will be linked to their trial registration number so that each trial is counted once, and the most complete analysis of each outcome will be used. Where a trial reports an outcome at several time points, the longest follow-up will be extracted, as decided in advance, to avoid choosing the most favourable result. Authors of trials with incomplete outcome data will be contacted once, and missing information will be recorded in the risk-of-bias assessment.
The example: risk of bias and synthesis
3.5 Risk of bias
Randomised trials will be assessed with the revised Cochrane tool, RoB 2, for each primary and secondary outcome (Sterne et al., 2019). Non-randomised studies contributing safety data will be assessed with ROBINS-I (Sterne et al., 2016). Assessments will be made independently by two reviewers.
3.6 Synthesis
Overall and progression-free survival will be pooled as log hazard ratios, and adverse events as risk ratios (Higgins et al., 2019), using random-effects meta-analysis because clinical variation between trials is expected. Heterogeneity will be described with the I² statistic and its confidence interval (Higgins et al., 2003) and explored in pre-specified subgroups (tumour type, line of therapy, monotherapy or combination). Where studies are too different to pool, results will be tabulated and synthesised narratively by outcome. Sensitivity analyses will exclude trials at high risk of bias.
3.7 Certainty of evidence
The certainty of evidence for each main outcome will be graded with the GRADE approach and presented in summary of findings tables (Guyatt et al., 2013). For each outcome these tables will show the number of trials and participants, the pooled effect with its 95% confidence interval, the corresponding absolute effect where baseline risks allow, and the certainty rating with the reasons for any downgrading. Before submission, the completed review will be checked against AMSTAR 2 as a self-assessment of methodological quality (Shea et al., 2017).
| Protocol element | Choice in this review | Why |
|---|---|---|
| Reporting guideline | PRISMA 2020 | Complete, comparable reporting (Page et al., 2021) |
| Registration | PROSPERO, before screening | Protects against changing methods after seeing data |
| Bias tool, trials | RoB 2 | Outcome-level assessment of randomised trials (Sterne et al., 2019) |
| Bias tool, non-randomised | ROBINS-I | Assesses confounding and selection (Sterne et al., 2016) |
| Effect measures | Hazard ratio; risk ratio | Suit time-to-event and binary outcomes (Higgins et al., 2019) |
| Certainty | GRADE | Transparent judgement of confidence (Guyatt et al., 2013) |
3.8 Ethics and dissemination
The review uses published and registry data only and involves no new participants, so formal ethical approval is not required; confirmation will be sought from the university ethics committee. The extracted data and analysis code will be shared in a public repository on publication, reviewers will declare any conflicts of interest, and results will be reported whatever their direction.
3.9 Timeline
| Months | Stage |
|---|---|
| 1 to 3 | Protocol development, piloting of search and forms, registration |
| 4 to 8 | Searches, de-duplication, title and abstract screening |
| 9 to 12 | Full-text screening, data extraction, risk-of-bias assessment |
| 13 to 16 | Meta-analysis, subgroup and sensitivity analyses, GRADE |
| 17 to 20 | Search update, writing and supervisor review |
How to adapt this to your own study
- Narrow your topic to a question you can state in PICO (or a suitable alternative) form.
- Register the protocol before screening and note every later change.
- Search several databases, trial registries and grey literature, and put the full strategy in an appendix.
- Name a risk-of-bias tool for each study design you include.
- Pre-specify effect measures, the meta-analysis model, heterogeneity assessment and subgroups.
- Grade the certainty of evidence for each main outcome.
- Check your references: avoid predatory sources and confirm that every DOI matches its citation.
Common mistakes in this chapter
- A title or question too broad to search systematically.
- No reporting guideline or registration.
- Naming "the Cochrane tool" without saying which one, or using one tool for every design.
- Restricting to English-language, published studies without justification.
- Promising a meta-analysis without saying when pooling is appropriate.
- Generic ethics and dissemination statements.
- Citing predatory or mismatched sources in the method itself.
| What examiners look for | Where this example does it | Check in your own draft |
|---|---|---|
| A focused, answerable question | Table 3.1 | Could two reviewers apply your criteria the same way? |
| Transparency and registration | 3.1 | Is your protocol registered and dated? |
| A comprehensive search | 3.3 and Appendix A | Have you searched beyond one or two databases? |
| Appropriate appraisal tools | 3.5 | Does each design have the right tool? |
| Pre-specified synthesis | 3.6 and Table 3.2 | Are effect measures and pooling rules fixed in advance? |
| Certainty of evidence | 3.7 | Will you say how confident readers can be in each result? |
| Reliable sources | References limited to verified methodological sources | Have you checked every reference? |
Key takeaways
- Frame a focused question and register the protocol before screening (Page et al., 2021).
- Search several databases, registries and grey literature (Hopewell et al., 2007).
- Match each study design to a named risk-of-bias tool (Sterne et al., 2019; Sterne et al., 2016).
- Fix effect measures and pooling rules in advance (Higgins et al., 2003).
- Grade certainty so readers know how far to trust each result (Guyatt et al., 2013).
